[Refactor] Clean up input preprocessing (#33687)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-02-05 18:43:42 +08:00
committed by GitHub
parent a2522839d8
commit 7bd42e609d
4 changed files with 101 additions and 214 deletions
+4
View File
@@ -1083,6 +1083,10 @@ class MultiModalEncDecInputs(MultiModalInputs):
Represents the outputs of
[`EncDecMultiModalProcessor`][vllm.multimodal.processing.EncDecMultiModalProcessor]
ready to be passed to vLLM internals.
Note: Even text-only encoder-decoder models are currently implemented
as multi-modal models for convenience.
(Example: https://github.com/neuralmagic/bart-plugin)
"""
encoder_prompt_token_ids: list[int]